DataCentersX > Stack > Cooling and Thermal Management > Cooling Water Systems
DC Cooling Water Systems
Water is the engineered substance that makes liquid cooling, direct-to-chip, and evaporative heat rejection possible. The pipes, pumps, CDUs, and cooling towers of a data center mechanical plant are inert without water flowing through them, and the quality of that water determines whether a cold plate survives ten years or fails in six months. This page covers the water systems themselves: the chemistry regimes in each loop, the makeup supply chain, the treatment chain that conditions incoming water to loop specifications, and the water balance that sets WUE for the facility as a whole.
The two-loop architecture established on the liquid cooling page provides the organizing frame. The facility water loop carries higher mineral content and tolerates open exposure to atmosphere through cooling towers. The technology cooling loop runs through microchannel cold plates and needs semiconductor-grade purity to avoid scaling, corrosion, and biofouling. Each loop has its own chemistry discipline and its own treatment infrastructure, and the facility's makeup water supply has to support both.
Water quality regimes
| Loop or Stream | Typical Conductivity | Key Controls | Failure Mode if Violated |
|---|---|---|---|
| Technology Cooling System (TCS) | Below 5 microsiemens per cm | Deionization, sub-micron filtration, pH, dissolved oxygen | Galvanic corrosion of mixed metals; microchannel fouling and blockage |
| Facility Water Loop (FWL) | Several hundred microsiemens per cm | Biocide, scale inhibitor, corrosion inhibitor, pH, blowdown cycles | Scale deposition, biofilm growth, tube corrosion |
| Cooling tower basin | Rises with cycles of concentration; blowdown limited | Cycles of concentration setpoint, biocide dosing, drift elimination | Legionella risk, scale on fill media, accelerated corrosion |
| Makeup water (incoming) | Varies by source; municipal typically 200 to 800 microsiemens | Source selection, softening, filtration, RO where needed | Contamination of downstream loops; treatment overload |
The technology cooling loop
The TCS is the most demanding water system in the facility and the one with the least margin for error. Water in this loop flows through microchannels in cold plates with hydraulic diameters in the 100 to 500 micron range. Any particulate larger than a fraction of that diameter can lodge in a channel and raise thermal resistance; enough such particulates can block flow entirely. Any dissolved mineral content that precipitates under operating conditions (higher temperature, changing pH) will deposit as scale on the channel walls. Any dissolved oxygen combined with mixed metallurgy will drive galvanic corrosion that degrades cold plate performance and releases metal ions into the loop.
The controls that keep TCS water within specification are continuous rather than periodic. Deionization beds remove dissolved ions and hold conductivity below the target. Sub-micron filtration (nominal 0.2 to 5 micron) captures particulates. Dissolved oxygen is controlled through degassing or scavenger chemistry. pH is bounded tightly, typically 8.0 to 9.5. Some operators add a corrosion inhibitor tuned to the specific metallurgy of their cold plates, manifolds, and fittings.
Glycol is added in climates where TCS piping runs through unconditioned space and freeze protection is required. Glycol concentrations of 20 to 40 percent are common in cold-climate facilities. The penalty is a small reduction in heat transfer performance and increased viscosity, which raises pumping power. The benefit is that the loop survives a power outage in winter without piping rupture.
The facility water loop
The FWL runs between the CDUs on the hall side and the mechanical plant on the rejection side. It carries higher mineral content than TCS because its exposure includes open cooling towers where evaporation concentrates solutes. The engineering discipline is less about purity and more about maintaining chemistry within the bounds where scale, corrosion, and biological growth are all simultaneously suppressed.
Scale inhibitors (typically phosphonate or polymer-based) prevent calcium and magnesium salts from precipitating on hot surfaces. Corrosion inhibitors protect the carbon steel, copper, and galvanized components common in facility piping. Biocides (oxidizing agents like chlorine dioxide, or non-oxidizing agents like isothiazolinones) suppress bacterial and algal growth that would otherwise foul heat exchangers and increase Legionella risk. Dosing is continuous and feedback-controlled against loop chemistry measurements.
Blowdown is the mechanism that keeps solute concentration bounded. As cooling tower water evaporates, pure water leaves and solutes stay behind. Blowdown periodically drains a fraction of the basin and replaces it with fresh makeup, resetting concentration to target. The ratio of makeup to blowdown is expressed as cycles of concentration: a tower running at five cycles of concentration passes five times as much makeup water through as it blows down. Higher cycles reduce water use but increase chemistry management burden; lower cycles simplify chemistry but waste water. Most hyperscale sites target four to seven cycles depending on makeup water quality.
Makeup water sources
Makeup water replaces what the facility consumes through evaporation, blowdown, drift, and leaks. The source of that makeup is a site-selection decision with long-running consequences for both facility economics and community relations.
| Source | Typical Quality | Constraints | Treatment Burden |
|---|---|---|---|
| Municipal potable | Treated, low variability, generally good | Withdrawal permits, reputational exposure, rate schedules | Light; softening and filtration usually sufficient |
| Groundwater (on-site wells) | Variable mineral content, potentially high hardness | Aquifer rights, drawdown impacts on neighbors, regulatory scrutiny | Moderate; softening, aeration, sometimes RO |
| Reclaimed/recycled municipal | Treated wastewater effluent; moderate solute load | Municipal infrastructure must exist and deliver to site | Moderate to heavy; additional filtration and chemistry adjustment |
| Non-potable surface | Variable; seasonal turbidity and biological load | Withdrawal permits, ecosystem impact assessment | Heavy; primary filtration, clarification, disinfection |
| Desalinated seawater | Pure after desalination; high energy cost per liter | Coastal site; energy price sensitivity | Light after desalination; RO plant is the main infrastructure |
Hyperscalers have increasingly moved toward reclaimed municipal water and non-potable sources for cooling makeup, preserving municipal potable supply for community use. This is as much a reputational and regulatory decision as an economic one: local opposition to large facilities drawing on potable water has grown in tandem with AI capacity buildouts, and reclaimed-water agreements provide both a stable supply and a defensible public position.
Treatment chain
Incoming makeup water passes through a treatment chain that conditions it to loop specifications. The chain varies by source quality and by which loops it feeds, but follows a consistent logical progression: remove suspended solids first, adjust mineral content second, polish to loop specifications last.
Primary filtration removes suspended solids through sand filters, cartridge filters, or multimedia beds. Softening ion-exchanges calcium and magnesium for sodium, reducing scaling potential in downstream equipment. Reverse osmosis rejects dissolved solids and produces high-purity water at the cost of brine concentrate that must be disposed. Deionization removes remaining ions to polish water to TCS specifications. Chemistry dosing adjusts pH, adds inhibitors, and maintains target biocide concentrations at the loop injection points.
The treatment chain serving the FWL is lighter than the chain serving the TCS, but both draw from the same makeup stream and the treatment plant has to produce water at both qualities. Hyperscale facilities typically include on-site water treatment as part of the mechanical plant rather than relying on municipal pretreatment, because the volume and quality requirements exceed what municipal systems deliver.
Water balance and WUE
Water Usage Effectiveness, WUE, measures liters of water consumed per kilowatt-hour of IT energy delivered. It is the water analog of PUE. The industry definition counts on-site water consumption only; a variant, source WUE, adds the water consumed at the power generation stage of the upstream electricity supply, which for thermoelectric generation is typically larger than the on-site number.
A facility's water balance is dominated by evaporative heat rejection. Tower evaporation accounts for the great majority of consumed water; blowdown and drift add smaller but real losses. Non-evaporative losses include closed-loop leaks, periodic system flushes during maintenance, and humidification of data halls running airside economization in dry climates. The design variable that most affects WUE is therefore the rejection technology choice, which returns to the wet-versus-dry tradeoff established on the cooling tower page.
A legacy evaporative hyperscale facility can run WUE in the 1.0 to 1.8 liters per kilowatt-hour range. A dry-cooled facility can approach zero on-site WUE at the cost of higher PUE. Hybrid facilities sit between, with seasonal water use that tracks ambient temperature. Sites in water-stressed regions are increasingly constrained by regulatory or permit limits on absolute water withdrawal, which forces hybrid or dry operation regardless of the PUE penalty.
Regulatory and community context
Water is the data center input facing the fastest-moving regulatory and community pressure. Electricity supply is contested through grid queue and interconnection timelines; water supply is contested through permitting, local ordinances, and public opposition. Several US jurisdictions have enacted water withdrawal caps on new data center builds; several European jurisdictions have tied data center permits to specific WUE commitments or to participation in reclaimed-water programs.
The structural implication for facility design is that the water system has to be defensible at three layers simultaneously: thermodynamically (meets loop chemistry specifications), economically (does not drive WUE above competitive benchmarks), and reputationally (does not impose perceived harm on the host community). Facilities that satisfy the first two but fail the third are increasingly blocked at the permit stage, regardless of engineering merit.
Related coverage
Cooling and Thermal Management | Cooling Tower and Heat Rejection | Liquid Cooling | Direct-to-Chip Cooling | Immersion Cooling | Facility Layer | Water Monitoring